The present invention relates to a semiconductor memory arrangement having a plurality of memory sectors which in turn comprise a plurality of storage locations, the storage locations in each memory sector being connectable via a local bit line to a global bit line and it being possible for the digital information stored at the storage locations to be read out and emitted via the appropriate global bit line by a sense amplifier.
A semiconductor memory arrangement of this kind having a hierarchical bit-line structure of this kind is known from EP 0 905 701 A2 for example. In this printed publication it is proposed that the storage locations of the semiconductor memory arrangement be arranged in rows and columns as a matrix, in which case local bit lines (LBL) in the columns are connected directly to the storage locations while a global bit line (GBL) is connected directly to a sense amplifier and, via suitable electronic switches, can be selectively connected to the appropriate local bit lines. To allow a storage location connected to a given local bit line to be read, the electronic switch which connects this local bit line to the global bit line is closed while the other switches in the relevant column of the semiconductor memory arrangement are opened. Each local bit line may for example be connected to several hundred storage locations.
For non-volatile memories, such as EEPROM memories (electrically erasable programmable read-only memories) or flash memories, there is a demand for high storage density, short access time, high operating frequency and low current consumption.
However, an increase in storage density involves an unwanted increase in the bit-line capacitance.
A prior-art semiconductor memory arrangement is shown by way of example in FIG. 2.
As is shown in FIG. 2, the semiconductor memory arrangement is divided into a plurality of memory sectors 6 which in turn contain a plurality of storage locations 8. Each storage location 8 is directly connected to a local bit line 5. Each local bit line 5 is connected to a corresponding global bit line 4 via a corresponding controllable switch 7, which is in the form of a switchable transistor in the present case. As shown in FIG. 2, the individual memory sectors 6 are arranged in columns and rows as a matrix, with a plurality of the global bit lines shown in FIG. 2 generally being provided as common to the memory sectors arranged in a column of the semiconductor memory arrangement, or rather to the storage locations 8 in the sectors. To allow data to be programmed to or read out from the individual storage locations 8, there is a central N-bit data bus 2 provided, it being possible for each of these N data-bus lines to be connected to a desired global bit line 4 via an electronically controllable switch 3, which is once again in the form of a switching transistor in the present case. In addition, each data-bus line in this N-bit data bus 2 is connected to a sense amplifier 1, thus enabling the N sense amplifiers to supply an N-bit output signal.
The individual local bit lines 5 define a local bit-line capacitance CLBL, while the individual global bit lines 4 define a global bit-line capacitance CGBL. In addition to this, the central read/write data bus 2 produces a bus capacitance CB. The total bit-line capacitance is substantially equal to the sum of the local bit-line capacitance CLBL, the global bit-line capacitance CGBL and the bus-line capacitance CB. The total bit-line capacitance depends on, amongst other things, the number of memory sectors 6 and their size. If it is assumed that, to increase storage density, it is merely the number of memory sectors that is increased, then this simply affects the global bit-line capacitance CGBL proportionally to the increase in the number of memory sectors 6. If on the other hand it is assumed that the size of the individual memory sectors 6 is increased as well, this results in an increase in the local bit-line capacitance CAL and bus-line capacitance CB too, because the size of the memory sectors 6, i.e. their length and width, affect the length of the global bit lines 5 and the length of the central bus line 2 to a corresponding degree.
The individual sense amplifiers 1 are among the most critical components of a non-volatile semiconductor memory and in particular are responsible for ensuring that a sufficiently high stability and read performance are maintained by the semiconductor memory. However, the increase in bit-line capacitance due to the circumstances described above contributes to an appreciable rise in access times, as also does that fact that as technology progresses there is a continuous fall in the current drawn by the location.
The object underlying the present invention is therefore to provide a semiconductor memory arrangement in which the bit-line capacitance is reduced without any appreciable increase in the area occupied.
This object is achieved in accordance with the invention by a semiconductor memory arrangement having the features detailed in claim 1. The subclaims each define preferred and advantageous embodiments of the present invention.
In accordance with the invention, the semiconductor memory arrangement is divided into a plurality of memory blocks, there being provided for each memory block no. i a group of mi sense amplifiers which is connected to a data bus associated with the given group and which can be connected in this way to corresponding global bit lines. The number of sense amplifiers in each group of sense amplifiers may be the same, in which case, to allow N bits to be read out, the number of data lines in the individual data buses and the number of sense amplifiers associated with each data bus is [[Mi=N/n]]mi=N/n (a being the number of memory blocks). In principle, it would of course also be conceivable for the number of sense amplifiers in each sense amplifier group, and hence the number of data lines in each data bus, i.e. the bit-width per bus, to differ from one another, all that needs to be ensured being that             ∑              i        =        1            n        ⁢          xe2x80x83        ⁢          m      i        =  N
Because of the division of the semiconductor memory arrangement as described above, the global bit-line capacitance can be appreciably reduced by reducing the number of memory sectors per memory block, assuming that the size of the sectors remains unchanged. The bus-line capacitance of the read/write data bus can also be appreciably reduced and there is no need in this case for, in particular, any change in the size of the individual memory sectors, because the bus-line capacitance is dependent simply on the length of a memory sector in the case of the solution according to the invention whereas in the prior art shown in FIG. 2 the bus-line capacitance depends on the length of at least two memory sectors arranged next to one another and on the space between the two memory sectors.
All in all, it is thus possible to achieve an appreciable reduction in the total bit-line capacitance with the help of the present invention, this also resulting in an appreciable reduction in current consumption because the current required to charge the bit-line capacitance, i.e. the charge required for this purpose per unit of time, is proportional to the bit-line capacitance. Also, no increase in the area occupied is required for the implementation of the present invention because the area needed for all the sense amplifiers in the case of the present invention is approximately equally as large as in the case of the prior art shown in FIG. 2.
For a 72 kB memory for example, with the sector size maintained, it has been possible to achieve a reduction in data-bus line capacitance of approx. 40-45% and hence a reduction in total bit-line capacitance of approx. 33%. At the same time, current consumption was lowered by approx. 100 xcexcA, which is a great advantage in low-power applications.
Although the present invention can preferably be used in non-volatile memories, and particularly in EEPROM or flash memories, the present invention is not of course limited to this preferred field of application and is independent of the type of memory used in the given case.